Hi 

I am implementing a DUC module in FPGA where QAM modulated 2400 symbols per second are continuously coming from a DSP. I wish to take these symbols to an IF of 455 KHZ sampled at 1 M from a DDS core. Do I have to interpolate the QAM symbols from 2400 symbols/sec to 1M symbols/sec as well for digital mixing? If so, then i am afraid that interpolation of QAM symbols might generate unwanted symbols such as 

x1 = { 1 5 3 10} might generate {1 2 5 4 3 6 10} if interpolated at twice the rate where symbol 2,4,6 are not present in orignal data. QAM modulation is implemented through look up table method.

Any kind of help in this regards is highly appreciated.

Regards

Mixing requires input and cos/sin to be at same sample rate.

You need to interpolate symbols properly(not repeat).

Az kaz mentioned, the mixer and signal sample rates need to match, so if you run the DDS at 1MHz and tune it accordingly the signal will need to be sampled at 1MHz.   In a transmitter this usually means some VERY careful interpolation/filtering so that adjacent channel energy is managed appropriately, i.e., you do not create or allow adjacent images or other energy to violate the required out-of-band transmit power spectral density.

It's all very doable, though.   This happens in most modulators.

Rereading your post it seems you are not even saying anything about pulse shaping. A classic modulator shapes pulses of symbols and upsamples to DAC speed then the receiver does the reverse. the pulses of symbols end up as peaks/dips after upsampling with each symbols occuping several samples but rceiver downsamples it back to final symbols.